Int. Journal of Math. Analysis, Vol. 7, 2013, no. 32, 1575-1586 HIKARI Ltd, www.m-hikari.com http://dx.doi.org/10.12988/ijma.2013.3359 On Bornological Divisors of Zero and Permanently Singular Elements in Multiplicative Convex Bornological Jordan Algebras Abdelaziz Tajmouati Sidi Mohamed Ben Abdellah University Faculty of Sciences Dhar El Marhaz Fez, Morocco abdelaziztajmouati@yahoo.fr Copyright c 2013 Abdelaziz Tajmouati. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Abstract In this paper we extended naturally the notion of topological divisors of zero (t.d.z) to the case of multiplicative convex bornological algebras. We prove that every element in topological boundary of the set of invertible elements is bornological divisor of Zero. On the other hand we give a characterization of bornological divisor of Zero, some results (Guelfand- Mazur,... ) are given. Keywords: Multiplicatively convex bornological algebras. Bornological Jordan algebras. Topological divisors of zero. Bornological divisors of Zero. Permanently singular elements 1 Introduction In [10] (see also [2]) we have already study the topological divisor of zero (t.d.z) in Jordan-Banach algebra. In this work we propose extend this notion to the case of multiplicative convex bornological Jordan algebras. The idea is to use the notion of topological divisor of zero in Jordan-Banach algebra, we have necessary to inject every sequence (x n ) n 0 in multiplicative convex bornological complete Jordan algebra to Jordan Banach algebra.this
1576 Abdelaziz Tajmouati is not true in general, the following example shows this Let a increasing sequence of Jordan Banach algebras (E n, n ) n 0, we suppose that the sequence ( n ) n 0 is increasing. Let E = n 0 E n,ine we take the bornology defined by: B E is bounded in E if, and only if, there exists n 1 such that B is bounded in E n. Then E is multiplicative convex bornological complete Jordan algebra. Let the sequence (x n ) n 0 such that, every n 0, x n E n and x n / E n 1. Then, (x n ) n 0 is not in E n. On other hand, if we consider a multiplicative convex bornological complete Jordan algebra of type M 1, then for every sequence (x n ) n 0 there exists a bounded B of E such that (x n ) n 0 E B. Indeed, let E = lime B multiplicative B convex bornological Jordan algebra of type M 1 and let (x n ) n 0 E. Then, for every n 0, there exists B n Bsuch that x n E Bn. Since E is the type M 1, there exists a sequence (λ n ) n 0 R + such that λ n B n is a bounded in E. n 0 then there is α R + and B Bsuch that λ n B n αb. Consequently, for n 0 every n 0, E Bn E λ nb n E B. This gives, (x n ) n 0 E B. n 0 Observe that every Jordan Banach algebra is a multiplicative convex bornological complete Jordan algebra of type M 1. Also, if E is unital topological Jordan algebra with continuous inverse such that it s F-space, then E is a multiplicative convex bornological complete Jordan algebra of type M 1. 2 Preliminaries Throughout this work, we suppose that the field K is either the real R or the complex field C. Recall that a bornology on a set X is a family B of subset X such that B is a covering of X, hereditary under inclusion and stable under finite union. The pair (X, B) is called bornological set. A subfamily B of B is said to be base of bornology B, if every element of B is contained in an element of B. A bornology B on K-vector space E is said to be vector bornology on E if the maps (x, y) x + y and (λ, x) λ.x are bounded. We called a bornological vector space(b.v.s) any pair (X, B) consisting of a vector space E and a vector bornology B on E.
On Bornological divisors 1577 A (b.v.s) space is called of type M 1 if, its satisfy the countability condition of Mackey: For every sequence of bounded (B k ) k in E, there exists a sequence of scalars (λ k ) k 0 such that λ k B k is bounded in E k=0 A vector bornology on a vector space is called a convex vector bornology if it is stable under the formation of convex hull. A bornological vector space is said a convex bornological vector space (cbvs) if its bornology is convex. A sequence (x n ) n 0 in bornological vector space (b.v.s) E is said Mackyconvergent to 0 (or converge bornological to 0) if there exists a bounded set B E such that ε >0 n 0 N, n n 0 implies x n εb. If E is (cbvs), then (x n ) n 0 is Mackey-convergent to 0 if there exists a bounded disk B E such that (x n ) n 0 E B and (x n ) n 0 converges to 0 in E B, where (E B,p B ) is the vector space spanned by B and endowed with the semi-norm p B gauge of B. A set B in a (bvs) E is said M-closed (or b-closed) if every sequence (x n ) n 0 B Macky convergent in E its limit belongs in B. Let E be a (cbvs) and A is a disk in E. A is called a completant disk if the space (E A,p A ) spanned by A and semi-normed by the gauge of A is a Banach space. A (cbvs) E is called a complete convex bornological vector space if its bornology has a base consisting of completant disks. For detail see [5], [6] and [7] An algebra over K is a K-vector space A with a bilinear map (x, y) x.y of A A into A. If this product is associative (resp commutative), we say that the algebra is associative (resp. commutative). Let x A, define the following map U x (y) =2x(xy) x 2 y, y A It is well-known that x U x is quadratic. An algebra A is a Jordan algebra if every (x, y) A 2 we have xy = yx and (x 2 y)x = x 2 y. Let A be an algebra, we denote by A + the algebra A equipped with its vector space structure and the product defined by : x y = 1 (xy + yx) for all 2 x, y A. The product is called the Jordan product. (A +, +,., ) is Jordan algebra. An normed Jordan algebra A is a Jordan algebra equipped by a norm satisfying xy x x for every (x, y) A 2,if(A, ) is Banach space then A is called a Jordan Banach algebra.
1578 Abdelaziz Tajmouati Let A be a Jordan Banach algebra with unit e, ifx A such that x e 1 then x is invertible. Let A be an algebra, Recall that a vector bornology B on A is called algebra bornology if the usual product is bounded in A A, in this case (A, B) is called bornological Jordan algebra. If again B have a pseudo basis B of idempotent disc, (A, B) is called convex multiplicative bornological Jordan algebra (cmbja). It is well known if (A, B) is (cmbja) then (A, B) is bornological inductive limit of semi normed Jordan algebras (A B, B B ) B B. Recall that a (cmbja) A is separated (respectively complete) if, and only if it s bornological inductive limit of normed Jordan algebras (respectively bornological inductive limit of Jordan Banach algebras). Let (A, B) a unital (cmbja) and x A, then x is invertible if, and only if, x is invertible in one A B which B B. If again (A, B) is complete then the group of invertible elements is Mackey open. Let A a unital Jordan algebra with unit e, then the Jacobson radical Rad(A) of A is defined by (For detail, see [4], [8], and [9]) Rad(A) ={x A : y A, e U x (y) is invertible } 3 Definitions and Properties Definition 3.1. Let (E,B) a complete (cmbja) and x E. x is said a bornological divisor of zero (bdz), if there exists B B such that x is a topological divisor of zero (tdz) in Jordan Banach algebra (E B,P B ). Definition 3.2. Let E be a Jordan Banach algebra and x E. x is called an absolute divisor of zero (a.d.z) if there is an element y 0of E such that U x (y) =0. For example every (a.d.z) in complete (cmbja) is a (bdz). Remark 1. In Jordan Banach algebra, the notion of (tdz) and (bdz) coincide. Proposition 3.1. Let (E,B) a complete (cmbja) and z some (bdz). Then, we have: 1) Fore every n 0, z n is a (bdz) 2) Fore every a E, U z (a) is a (bdz)
On Bornological divisors 1579 Proof: Applique [10, proposition 5.1, and proposition 5.2] Proposition 3.2. Let (E,B) an unital complete (cmbja) of type M 1. Then. Every non invertible element which is limit on Mackey sense of a sequence of invertible elements in E is a (bdz). On other expression, Fr(G(E)) Z b, where G(E) is a groupe of invertible elements of E, Fr is the topological boundary of M-closeness and Z b the set of a (bdz) of E. For proof, we need the necessary following lemma. Lemma 3.1. Let (E,B) a unital complete (cmbja) of type M 1.Let(x n ) n 0 a sequence of invertible elements in E which converges of Mackey-sene to x. If (x 2 n ) n 0 is bounded, then x is invertible element in E. Proof: Then Let e the uint of E. we have ( n 0),U xn (x 2 n )=e U x (x 2 n ) e =(U x U xn )(x 2 n ) Since (x 2 n ) n 0 is bounded, there exists α > 0 and B 1 B such that (x 2 n ) n 0 αb 1. On other hand (x n ) n 0 converges of Mackey sense to an element x in E, then there exists B 2 Bsuch that (x n ) n 0 E B2,x E B2 and lim p B 2 (x n x) =0. B is growth filtrating, then there exists B Bsuch that: (x 2 n ) n 0 E B, (x n ) n 0 E B,x E B and lim p B (x n x) =0. Since z U z is continuous in E B, then lim U xn U x =0. Consequently we have: n 0 N n N, n N n U xn U x 1. n 0 N n N, n N n and p B (t) 1 p B (U xn (t) U x (t)) 1. Since for every n, p B ( 1 α x 2 n ) 1 then for n sufficiently large
1580 Abdelaziz Tajmouati p B (U xn ( 1 α x 2 n ) U x( 1 α x 2 n )) 1 implies p B(U x ( 1 α x 2 n ) e) 1. Then, for n sufficiently large, U x ( 1 α x 2 n ) is invertible in E B. Consequently, x is invertible in E. Proof of proposition Let y n the inverse of x n for all n. (yn) 2 n is not bounded in E (Lemma3.1). Then, it s not bounded in every (E B,p B ).Ebeing of type M 1, then there exists B 1 Bsuch that (yn) 2 n E B1. On other hand, (x n ) n 0 bornological converges to x, then there is B 2 Bsuch that (x n ) n 0 and x belongs in E B2 and lim p B2 (x n x) =0. B being growth filtrating, then there exists B Bsuch that (yn 2) n E B, (x n ) n 0 E B,x E B and lim p B (x n x) =0. We can suppose that, for every n 0, yn 2 0 and lim p B (y 2 n)=0. Put z n = y2 n p B (yn 2 ). We have z n =1, for all n 0 and U z (z n )= e +(U p B (yn 2 ) z U xn )(z n ). Since p B [(U z U xn )(z n )] U z U xn, we obtain lim U z (z n )=0. Then, z is a (tdz) in (E B,p B ). Consequently, z is a (bdz) in E. If [G(E)] (1) is the closure of G(E) for the M-closeness topological of E we have [G(E)] (1) [G(E)] c Z b. Since G(E) is M-open. Then Fr(G(E)) Z b. From the above proposition we immediately obtain the followings corollary, which generalize some results in [8]. Corollary 3.1. Let (E,B) an unital complete (cmbja) of type M 1. Let x E, then for every scalar λ in bornological boundary of sp E (x) the element x λe is a (bdz). Proof: Since sp E (x) is non-empty closed in C, therfore Fr(sp E (x)) = sp E (x) ([sp E (x)]) c Let λ Fr(sp E (x)), then x λe is non-invertible in E and there exists (λ n ) n C converging to λ such that x λ n e is invertible in E for each n 0.
On Bornological divisors 1581 Hence (x λ n e) n converges in Mackey sense to x λe. Consequently, x is a (bdz). Corollary 3.2. Let (E,B) an unital complete (cmbja) of type M 1. Every element in the Jacobson radical of E is a (bdz). Proof: Let x Rad(E), then sp E (x) ={0}. Hence Fr(sp E (x)) = {0}. So that x a (bdz). Corollary 3.3. Let (E,B) an unital complete (cmbja) of type M 1. Let F be a closed subalgebra of E such that F contains the unit of E, then for every x F we have: sp E (x) sp F (x) and Fr(sp F (x)) Fr(sp E (x)) Corollary 3.4. Let (E,B) an unital complete (cmbja) of type M 1. If E has not a non-zero (bdz), then E is bornological isomorphic to C. Proof: Let x E\{0} and λ Fr(sp E (x)). Then x λe is a (bdz), hence x = λe, this proves the corollary. The characterization of (bdz) need the following definition. Definition 3.3. Let (E,B) a separated (cmbja). For every B in B and x E B, denotes by λ B (x) = inf z E B p B (z) =1 p B (U x (z)) The following proposition characterize the (bdz). Proposition 3.3. Let (E,B) a complete (cmbja) and x E. x is a (bdz) if, and only if, there exists B Bsuch that λ B (x) =0. Proof: It suffice to applique [10, definition 3.1] and the definition of (bdz) and the characterization of the infimum. Proposition 3.4. Let (E,B) an unital complete (cmbja) of type M 1. Suppose that there exists a constant K R + such that B B p B (U x (y)) Kp B (x)p B (y) x, y E B. Then, E is bornological isomorphic to C.
1582 Abdelaziz Tajmouati Proof: It suffices to show that E haven t any non-zero (bdz). Suppose that there is z E such that is a (bdz), then there exists B Bfor which z is a (dtz) in (E B,p B ). Hence λ B (z) =0. Since for every y E we have λ B (z) Kp B (z). It follows p B (z) =0, consequently z = 0, and therefore E haven t any non-zero (bdz). This proves that E is bornological isomorphic to C. Definition 3.4. Let E a separated (cmbja). An element x E is said to satisfy the condition (D.F) if there exists a non-bounded sequence (x n ) n 0 in E such that (U x (x n )) n 0 converges in the Mackey-sense to 0. Proposition 3.5. Let (E,B) an unital complete (cmbja) of type M 1. Then Every element x E satisfying the condition (D.F) is a (bdz). Proof: Suppose that x E satisfies the condition (D.F), then there exists B 0 Bsuch that x E B0 and a non-bounded sequence (x n ) n 0 E where (U x (x n )) n 0 converges in the Mackey-sense to 0. Therefore, there exists B 1 B such that (U x (x n )) n 0 E B1 converges to 0 in (E B1,p B1 ). Since E is of type M 1, then there exists B 2 Bsuch that (x n ) n 0 E B2. B being growth filtrating, then there exists B Band x E B, (x n 0 ) n E B, (U x (x n )) n 0 E B, (U x (x n 0 )) n converges to 0 in (E B,p B ) and (x n ) n 0 is nt bounded in (E B,p B ). Therefore, (x n ) n 0 not converging to 0 in (E B,p B ). From which we conclude that x is a (bdz) in E. A natural question is given, what can we say for the conversely of the proposition 3.5? Unfortunately, in general situation, if x is a (bdz) in unital complete (cmbja) of type M 1, then x can not satisfies the condition (D.F) as the following example shows Let A the set of all real or complex sequences. Let x =(x n ) n 0 and y =(y n ) n 0 two elements in A and λ C. In A we defined the following usual operations : x + y =(x n + y n ) n, λ.x =(λx n ) n and x y =(z n ) n where z n = n x p y n p. With this operations, A is an unital commutative algebra with unit e =(e n ) n 0 such that e 0 = 1 and e n =0 n 1. For all n 0, put p n (x) = n x i where x =(x i ) i 0. i=0 p n is a sub-multiplicative semi-norm on A. p=0
On Bornological divisors 1583 Indeed, let x =(x n ) n 0 and y =(y n ) n 0 in A. x y =(z n ) n such that z n = n x p y n p. p=0 Let n 0, p n (x y) = n z i = z 0 + z 1 +... + z n i=0, = x 0 y 0 + x 0 y 1 + x 1 y 0 +... + x 0 y n + x 1 y n 1 +... + x n y 0 ( n x i )( n y i )=p n (x)p n (y). i=0 i=0 With the sequence of semi-norms (p n ) n 0,Ais an unital commutative metrizable complete (cmbja). On other hand, the group of invertible elements of A is exactly {x =(x n ) n 0 A : x 0 0}, which is open in A. Then, A is an unital commutative Frechet (cmba) which is Q-algebra. Hence, every element of A is regular, consequently, A is unital commutative (cmbja) complete for Von-Neumann bornology [1]. Since A is metrizable, then it s of the type M 1 [ 7,p.226]. Now, Consider z =(z n ) n 1 such that z 1 = 1 and z n =0, n 1. We show that z is a (bdz). Of course, it suffices to prove that z belongs in the Jacobson radical of A. Let y =(y n ) n 0 A, the term of row zero in sequence e + z y is 1. Hence e + z y is invertible in A. Therefore z belongs in radical of A. It then follows that z is a (bdz). Next, we prove that for every non-bounded (x m ) m 0 A, the sequence (z x m ) m 0 not converges to 0 in Mackey-sense. For all m 0, set x m =(x m n ) n 0. z x m =(z n ) n 0 +(x m n ) n 0 =(t m n ), with t m 0 = 0 and t m n = x m n, ( n 0). Then, the sequence (z x m ) m 0 is not bounded. Consequently, (z x m ) m 0 not converges to 0 in Mackey -sense. Observe that for Jordan Banach algebra we have the following characterization. Proposition 3.6. Let E be an Jordan Banach algebra. Let x be a given element in E. Then x is a (tdz) if, and only if, x satisfies the condition (D.F). Proof: Let x E, and suppose that there exists a non-bounded sequence (x n ) n 0 in E such that (U x (x n )) n 0 converges bornological to 0. Then, (U x (x n )) n 0 converges topologically to 0 and (x n ) n 0 not converges to 0 in E. Therefore, x is a (tdz) in E. Conversely, if x is a (tdz) in E, there exists a sequence (x n ) n 0 in E such that
1584 Abdelaziz Tajmouati (U x (x n )) n 0 converges to 0 and (x n ) n 0 not converges to 0. The map (x, y) x y is a distance invariant by translation and defines the topology of E. hence, there exists a sequence (γ n ) n 0 R + such that lim γ n =+ and lim γ n U x (x n )=0. For all n 0 put y n = γ n x n. Then, (y n ) n 0 is not bounded and (U x (y n )) n 0 converges bornologically to 0 [5]. Proposition 3.7. Let E = lim B E B be a complete (cmbja). Then Every absolute divisor of zero (a.d.z) satisfies the condition (D.F). Proof: Let x a (a.d.z). Then there exists y 0 such that U x (y) =0. For every n 0, put y n = ny. Then (y n ) n 0 is not bounded in E and U x (y n )=0, for all n 0. Therefore, (U x (y n )) n 0 converges bornological to 0. 4 Permanently singular elements in complete multiplicative convex bornological Jordan algebras Definition 4.1. Let E an unital separated (cmbja). An unital separated (cmbja) E 1 is called an extension of E, ife is subalgebra of E 1, the unit element of E 1 is a unit element for E and the induced bornology of E 1 on E coincide with the bornology of E. Definition 4.2. Let E an unital separated (cmbja). An element in E is called permanently singular if it s not invertible in every extension of E. Proposition 4.1. Let E an unital complete (cmbja). Every element in E satisfying the condition (D.F) is not invertible in E. Proof: Let x E where E = lime B satisfying the condition (D.F). B Then, there exists a non-bounded sequence (x n ) n 0 E such that (U x (x n )) n 0 converges bornological to 0. Then, there is B 0 B such that (U x (x n )) n 0 converges to 0 in (E B0,p B0 ). Since B is growth filtrating, we can suppose that (x n ) n 0 E B0. Suppose that x is invertible in E. Then, there is B 1 Bsuch that x is invertible in E B1. Then, there is B Bwhere E B0 E B,E B1 E B, p B p B1, (x n ) n 0 E B,(U x (x n )) n 0 E B and lim p B (U x (x n )) = 0. On other hand x is invertible in E B, hence U x is invertible in L(E B ) and we have U x 1 =(U x ) 1, the map z U x 1(z) is continuous and
On Bornological divisors 1585 p B (x n )=p B [U x 1(U x (x n ))], then lim p B (x n )=0. We get a contradiction. Hence, x is not invertible in E. The proof is complete. Proposition 4.2. Let E an unital complete (cmbja). Let x E, ifx satisfies the condition (D.F) in E, then x is permanently singular in E. Proof: We can prove that if x satisfies the condition (D.F) ine. Then it satisfies this condition in every extension of E. Consequently, x is permanently singular in E. Proposition 4.3. Let E an unital complete (cmbja). Let x E, ifx satisfies the condition (D.F) in E and if y E, then again the element U x (y) satisfies the condition (D.F). Proof: For two elements x, y in E, we have U Ux(y) = U x U y U x. If x satisfies the condition (D.F) ine, then there exists a non-bounded sequence (x n ) n 0 E such that (U x (x n )) n 0 converges bornological to 0. Hence there is B B such that (U x (x n )) n 0 converges to 0 in (E B,p B ). Hence the map z U x U y (z) is continuous in E B, the sequence (U x U y (U x (x n ))) n 0 converges to 0 in E B. Therefore, (U Ux(y)(x n )) n 0 converges bornological to 0. We conclude that U x (y) satisfies the condition (D.F). References [1] M. AKKAR. Etude spectrale et structure d algebres topologiques et bornologiques completes. These de doctorat d Etat SC. Math. Universite Bordeaux I, 1976. [2] G.R.ALLAN. A spectral theorie for loclly convex algebras. Proc. Lond. Soc,3. 15, 1965, 399-421. [3] C.V.DEVAPAKKIAM. Jordan Algebras with continuous inverse. Math. Japon. 16 (1971), pp. 115-125. [4] HARALD and ERLINNG. Jordan operator algebras. Pitman Advanced. Publicing program. Boston. [5] H. HOGBE-NLEND. Theorie des bornologies et applications. Lecture Notes 213, Springer-Verlag, 1971. [6] H. HOGBE-NLEND. Les fondements de la theorie spectrale des algebres bornologiques., Separata Do Boletim Da Sociedade Brasileira De Mathematica. Vol. 3, 1972
1586 Abdelaziz Tajmouati [7] H. HOGBE-NLEND. Un cours introductif sur les bornologies et l analyse fonctionnelle., Universite Bordeaux I, 1972. [8] N. JACOBSON. Structure and representations of Jordan algebras., American Mathematical Society, Providence, 1968. [9] J. MARTNEZ MORINO. Sobre algebras de Jordan normadas completas, Tesis doctoral, universidad de Granada, 1977. [10] A. TAJMOUATI. On Topological Divisor of Zero and Permanently Singular Elements in Jordan Banach Algebras. Int. Journal of Math. Analysis, Vol.6, 2012, 58, pp. 2871-2884. Received: March 12, 2013